DENSE MEDIA SEPARATION METHOD

Field of the Invention

The present invention relates to the separation of solids. The invention relates particularly to fine (less than 1 mm) feed Dense Media Separation (DMS).

Background to the Invention

Dense Media Separation (DMS) is a process widely used in the mining industry to separate the valuable minerals from the non-valuable rock by differences in density. For example, DMS can be used in the iron ore industry because hematite is denser than quartz, muscovite and kaolinite in general.

The DMS process involves the use of a suspension of particulate material in a liquid, typically water. The particulate material, or media preferably comprises magnetic particles, for example magnetite or ferrosilicon (FeSi) particles because this facilitates the recovery of the particulate material for reuse after the separation process. The particles of the particulate material are sufficiently fine to allow their stable suspension in the relevant liquid, and typically take the form of powder, while being sufficiently dense/heavy to provide the required media density.

The media particles are typically formed by milling or atomisation. The resulting media suspension is commonly referred to as a dense medium. Where the particulate material comprises magnetic or magnetised particles, the media suspension has a density greater than that of the liquid alone. For example a typical dense medium may have an apparent density of, say, 2.65 specific gravity while the specific gravity of water is 1 . The advantage of using a magnetic particulate material is to facilitate subsequent retrieval of the particulate material for reuse. During use, the media suspension is contained in a separation vessel, for example a cyclone vessel (sometimes referred to as a dense medium cyclone). The media suspension is usually mixed with the solids to be separated (typically comprising ore) before being transferred to the separation vessel. Where the separation vessel comprises a cyclone, separation is effected by differences in centrifugal force experienced by particles of the solids to be separated of differing density, the less dense material tending to float in the liquid suspension and so exiting the cyclone at the top (“floats”), while the denser material sinks and exits through the bottom (“sinks”).

A problem with DMS is that the suspended media tends to separate from the media suspension along with the solids to be separated as a result of its relatively high density (typically between 6.7 and 7.1 specific gravity for ferrosilicon). Therefore a stable media is required for optimum DMS efficiencies, and optimum efficiencies are a priority more than ever with high commodity prices. Stability is achieved using powdered media that is fine enough to prevent rapid settling of the media under the centrifugal forces in the cyclone or gravity in the case of a Dense Media Drum. It is this fineness that gives rise to most media losses for reasons including the following:

1 . Fine suspension media adheres to the ore/solids surface and is difficult to wash off from the recovered product at the end of the process. 2. Fine suspension media is more susceptible to corrosion (e.g. oxidation) due to the high surface area to volume.

3. Fine suspension media is more difficult to recover in magnetic separators.

The higher hydrodynamic drag forces that fine particles experience, results in poor recovery of finer media in the magnetic separators.

Commercially available Ferrosilicon is manufactured as either milled or atomised. The atomised version is commonly manufactured in five size fractions: Special Coarse, Coarse, Fine, Cyclone 60 and Cyclone 40 and, because it is spherical, it is more easily washed, more resistant to corrosion but is more expensive. Milled ferrosilicon is cheaper and is commercially available in six different sizes: 1 00#, 65D, 1 00D, 1 50D, 270D, 270F (from for example DMS Powders (www.dmspowders.com) or M & M Alloys Limited (www.mandmalloys.com). In conventional DMS plants where the required media specific gravity is greater than 3.2, as in iron ores, the viscosity of the milled media is too great for efficient separation and atomised ferrosilicon is used.

Generally, the smaller the cyclone diameter the easier it is to operate at higher feed pressures which results in larger centrifugal forces experienced by the media and ore particles in the cyclone.

Typically, ferrosilicon losses in cyclone DMS circuits range from 1 20g ferrosilicon per tonne (g/t) up to 500 g/t. Media losses are known to represent from 20% to 40% of the total operating costs of a DMS plant.

It would be desirable to improve the separation efficiency for -1 mm feed particle size solid feeds. It is further desirable to the reduce media losses for -1 mm feed DMS systems.

Summary of the invention

The invention provides a method of separating solids, the method comprising: i) adding the solids to a suspension of particulate material comprising magnetic particles in a liquid to create a mixture, ii) locating the mixture in a separation vessel such that rotation is imparted to the mixture around a space bounded by an outer wall of the vessel to impart a centrifugal force on the solids; iii) the mixture is separated into a floats and sinks fractions;

wherein said particulate material is ferrosilicon having D90 particle size of below 200pm,

the solids is iron ore having a particle size of less than 1 mm and greater than 200 micron (-1 mm +200 pm), for example (- 1 mm+212pm), preferably greater than 400 micron (pm) (- 1 +400pm),

the feed density of the mixture is less than 3.8 g/cm ³ , preferably less than 3.6 g/cm ³ ,

the operating pressure of the mixture is from 10 to 1 5D,

the cut point differential is between 0.1 and 0.6 of the feed density, and

wherein the mixture includes a slimes iron ore fraction of less than 1 0%.wt. having a particle size of less than 45pm.

The suspension of particulate material comprising magnetic particles in a liquid may be known as the medium. In one embodiment the floats and sinks fractions are subject to magnetic separation before washing to recover the particulate material.

The floats and sinks fractions may be passed over separation screens vibrating at a frequency of greater than about 50Hz. The separation screens may have a double oscillation.

Detailed Description of the Invention

In this specification cyclone Operating Head is a function of Feed Pressure and Feed Density (and is reported as a ratio of Total Head (m) and

Cyclone Diameter (m), hence the use of“D”.

Operating Head (D) = ((Feed Pressure (kPa) / (Feed Density (t/m ³ ) X Gravitational Acceleration (m/s ² ) )) / (Diameter of a DMS Cyclone(m))

Operating Head and Operating Pressure are often used interchangeably to mean the same thing.

Testwork has previously been conducted on samples from Kumba Iron Ore to compare the performance of DMS to that of a spiral circuit on fine iron ore

(-1 mm+212um), using both high and low grade material from Sishen mine.

It was found that the density distribution of the samples evaluated played an important role in the efficiency of separation (Ep) and hence the overall grades and recoveries obtained. High amounts of near-density material were shown to have a negative effect on both gravity separation methods. It was also demonstrated that the size distribution of the material is critical in determining the overall Ep obtained in a DMS plant, with a significant deterioration of Ep with a reduction in particle size. A key finding was that when operating at low DMS feed densities of around 3.4g/cm ³ , Cyclone 60 FeSi was able to obtain outpoints of >4.0g/cm ³ on the fine iron ore (<1 mm) tested. The use of the more expensive gas atomized Exxaro FeSi was not recommended at that stage, as the required outpoint on the DMS plant could be obtained with the Cyclone 60 FeSi. Note , however, that both these batch DMS tests were done without any slimes build-up in the circuit as would be expected in a continuously operating plant. A build-up of slimes is expected to significantly alter the rheology inside the DMS cyclone along with the cyclone performance efficiency.

The current objective of the testwork was therefore to compare the performance of both atomised FeSi types across a range of operating conditions in the presence of slimes build-up, as well as to define DMS modeling parameters for each of the FeSi types.

Anglo American Kumba Iron Ore supplied two bulk samples of <1 mm ore for test work on the pilot DMS plant:

1 . low grade (BIF sample crushed to <1 mm using FIPGR) and

2. high grade iron ore (Jig plant feed fines, conventionally crushed to < 1 mm).

Experimental Procedures

Sample preparation

Feed preparation

Approximately 1 .5 tons of low grade Banded Iron Stone Formation (BIF) and 4.5 tons of high grade Jig (-1 mm) was provided for the testwork. The following sample preparation was conducted on each sample as received:

• The samples were wet screened at 212pm to produce DMS feed at <1 mm >212pm. • The -212 m fraction was further screened at 45pm to produce a <45pm fraction to be used as slimes addition during the DMS runs.

The following feed characterisation was conducted on -1 mm +212pm sub samples:

• Particle size distribution (PSD) was done on a 2x1 kg sub-sample, using root two sizing.

• Head chemical analyses were conducted on 2x1 OOg sub-samples in duplicate to ascertain head grade of the sample.

• A 1 kg sub-sample was sized into -1 mm+600um and -600+212um.

These fractions were submitted to Mineralogy for Particle Tracking Analysis (PTA) analysis, using MLA.

• A 500g sub-sample were submitted to mineralogy for electron microprobe analysis (EMPA)

• Sub-samples of 1 00kg each (-1 mm+212pm) BI F (low grade) and Jig feed fines (high grade) were prepared for pilot DMS testwork. Individual 1 00kg samples were used for batch DMS runs at feed relative densities of 3.4, 3.6, 3.8 and 4.0 g/cm ³ with introduction of - 45pm slimes at 5% by mass of the medium. Some of the tests were conducted without introduction of -45pm slimes for comparison.

DMS Testwork Products

During DMS operation, bulk mass of sink and floats were taken. The mass splits across the DMS plant were determined and sub-samples submitted for chemical analysis. Sub-samples of each product were also submitted for PTA analysis, using MLA.

Scope of work

Test work strategy Figure 1 and Table 1 below outlines the test work strategy for the DMS runs. In total, 1 6 DMS tests were conducted - an initial 1 1 tests followed by an additional 5 tests.

Table 1 : Summary of tests conducted

Mineralogy - Particle Tracking Analysis

Mineralogy (Particle Tracking Analysis - PTA) was used to characterize each ore type in terms of mineral, density and size distribution of the feed and DMS products. The liberation characteristics of each mineral by size class were also reported.

PTA was conducted on the two feed sub-samples as well as on all the products of the initial 1 1 tests. Due to the coarse nature of the particles within the samples, each size fraction submitted was mounted into slabs, carbon coated and analysed using autoSEM (MLA). During measurement, the MLA generates an X-ray analysis for each region (grey level) within a particle. The measurement mode employed in this study was chosen on the basis of the ore type, run time, purpose of study, particle size and the successful ability to delineate mineral grain boundaries in particles. During the investigation, between 2500 and 20000 particles per polished section for each fraction were analysed and processed via AutoSEM. A larger number of particles aid in the acquisition of a statistically representative dataset of the overall sample. Particle characterisation data pertaining to mineral types/compositions, particle size, density, weight percent of the particle population, area of particle, shape factor, circularity, and perimeter of each particle were ascertained during offline processing. For purposes of this study, calculations based on shape descriptors such as shape factor, circularity and equivalent circle diameter (ECD) were included in the overall data set for each size fraction.

Pilot Dense Medium Separation (DMS) Tests

Figure 2 and Figure 3 shows the pilot DMS test rig flowsheet and DMS pilot plant setup respectively used for the testwork. The DMS plant is equipped with a 250mm diameter Multotec cyclone. The pilot test operating parameters were monitored during the test. These include circulating feed density, cyclone underflow and overflow densities to ensure steady state conditions. Feed pressure was also monitored to ensure constant feed pressure to the cyclone. The DMS sieve bands of 600mGP were covered with wire mesh of 250pm to help avoid fines going through sieve bands. The DMS testwork was conducted to target above 1 2D values at different feed densities. A density differential less than 0.4 RD is recommended for the efficient operation of the cyclone when the proportion of near gravity feed material is high. The unit has no moving parts which reduces the maintenance requirements.

Once the required parameters were reached and the steady state conditions established, the float and sinks products from each run were collected over a period of 30-45 minutes, sun dried and weighed. The products were screened using 212pm screen to remove FeSi and a magnet was used to further remove residual FeSi on the +212pm fractions. Representative 7 sub-samples of 1 0kg from both float and sinks of each run were sub-sampled and the products were sent for chemical as well as mineralogical analysis.

Chemical Analysis

All samples were submitted for chemical analysis. Sample preparation was done, using a tungsten-carbide pulverising bowl, which was cleaned at intervals, using soaped water and acetone, to replace the conventional cleaning by silica sand as silica is the main contaminant. A standard ISO 1 7025 accredited hematite XRF assay technique was used to determine Fe and major gangue elements. Table 2 below show the description of the method used to analyse the ore products for the testwork conducted.

Table 2: Analytical method used for ore analysis

Modelling

The PTA results obtained from all the DMS tests conducted were used to determine the parameters of a standard DMS Weibull model for each of the FeSi types. The model takes the mineral, density and size distribution of the feed material into account to predict DMS performance, once the parameters are defined. Correct characterisation of the feed material is essential in order to predict performance with confidence.

Y = 100(1 -exp(-(ln(1/(1 - Yp))) ^* (D/Rhop) ^A (p ^* d ^A q)))

. Equation 1

Also included in the Weibull model are the geometric mean particle size (d) and the mean particle density (D). Sets of values for these parameters were derived for each individual DMS run conducted.

RESULTS

Feed Characterisation Feed Size Distribution

The samples received were wet screened separately, using a 21 2pm Sweco screen. Subsequent to wet screening, approximately 64% of BIF (low grade) and 55% of the Jig feed fines (High grade) reported to the - 1 mm+21 2pm fraction. A root 2 size distribution of the feed material was conducted in duplicate.

For each sample, a PTA was conducted on the feed as well as the DMS sinks and floats from each tests. The DMS sinks and floats results were used to construct a reconstituted feed size distribution that could be compared to the measured feed. These results are all reported in Figure 4 and 5 for -1 mm+21 2pm, BIF (low grade) and Jig feed fines (High grade) samples respectively.

In terms of particle size distribution, the duplicate results were reproducible. The PTA analysis produced a slightly finer size distribution of the feed on average (based on the reconstituted results). This outcome is related to the well-known stereological effect created from 2D sectioning of 3D particles for MLA measurements. The total head grade of the low grade sample was 39.7% Fe and that of the Jig sample was 48.2% Fe. The detailed results are reported in Appendix A.

Electron Microprobe Analysis (EMPA) of Flematite particles

BIF (Low grade) sample

Mineralogical evaluation of the BIF (low grade) sample indicated that the hematite particles did not contain porosity or micro-inclusions. The hematite was thus classified as high density, high Fe grade particles and EMPA was thus not conducted on this sample.

Jig Fines Feed (High Grade) sample

EMPA was conducted on the high grade sample, in order to classify the hematite phases correctly and then reliably assign an Fe -grade and density to each phase identified. Figure 6 shows a SEM image of the high grade sample. Note that the number within the grains shows the sites of individual EPMA measurements.

The dark grey grains in the above image were classified as hematite A and consist of clean high-quality hematite. The medium gray grains represent clean high-quality hematite containing micropores and were classified as hematite B. The presence of the micropores resulted in a lower average relative density for this hematite class which will be realistically experienced in water based separation. The lower density will not be evident when pycnometry is done, since the particles will be pulverised and the pores removed. The light gray grains were classified as “hematite” C and are inherently low-quality hematites which contain both micro pores and micro-inclusions of Si, Al and K bearing minerals. These inclusions are so fine (d Omicron) that practical liberation will not be possible. Note that a wide range of composition for this class was observed by EPMA and the Fe content and relative densities assigned to this class are averages only. The EMPA data collected is detailed in Appendix B.

Mineral Reference list

Subsequent to defining the hematite species identified during the EMPA, a minerals reference list is defined, which is used by the MLA, to classify the particles identified correctly. Tables 3 and Table 4 show the mineral reference lists for the BIF (low grade) and Jig Fines Feed (high grade) samples respectively.

Table 3 shows only one phase of hematite (A) being present in the low grade sample, with a pure hematite content reported as 66.8% Fe at a density of 5.05g/cm ³ . Three hematite phases (Table 4) were observed in the high grade sample with hematite classes ranging from 62.0% to 67.2% and densities between 4.5-5.1 g/cm ³ . Table 3: Mineral reference list for low grade sample

Table 4: Mineral reference list for high grade sample

Feed Mineral distribution and liberation

The PTA results obtained were used to characterise the feed in terms of size, density and mineral distribution. Liberation characteristics of hematite and gangue were also defined. Table 5 and Table 6 show the mineral distribution and liberation of the low grade and high grade sample respectively. Both the tables and Figure 7 shows that 35.7% of hematite A is totally liberated in low grade sample and 60.6% of hematite A is totally liberated in high grade sample. Flematite B (porous) in the high grade sample is locked, with only 17.7% reporting as totally liberated. Flematite C (porous, with micro-inclusions), although low grade, has a reasonably high proportion of totally liberated material at 55.2%. Table 5: Mineral distribution and liberation of low grade sample

Table 6: Mineral distribution and liberation of high grade sample

Table 6 shows a significant portion of hematite C (26.3%) present in the high grade sample. This will contribute to a high amount of near-density material of the DMS feed as well as negatively affect the attainable product Fe grade.

Figure 7 comparatively summarises hematite liberation for the three phases identified, for both the high grade and low grade sample.

Density distribution Figures 8 to 1 1 show the density by size distribution of the hematite types present in the low grade and high grade samples. These figures attempt to visually represent the contribution of each size class of DMS feed, highlighting the mass percent and amount of near-density fractions present.

The cumulative density and grade distributions reported in Table 7 shows the best theoretical yield-grade performance possible for both low grade and high grade samples. The tables show that, under ideal separation conditions:

• The low grade sample, cut at an s.g. of 4.0 would yield 51 .8% product at 64.1 % Fe

• The high grade sample, cut at an s.g. of 4.1 would yield 67.4% product at 64.1 % Fe

(Note: the FIG Fe grade in the density class 4.9g/ml is given as 62.66%Fe. This is due to the material in this class being dominated by Flematite C (1 1 .6% mass) locked with Flematite A (3.5% mass)).

Detailed size and density characterisation results are presented in Appendix C.

Table 7: Feed density distribution for low grade and high grade

samples

Dense Medium Separation (DMS) Testwork Testwork conditions

DMS testwork was conducted on a pilot plant equipped with a 250mm DMS Multotec cyclone. The -1 mm +21 2pm batches prepared from each ore type were run on the DMS plant, using Cyclone 60 and Exxaro FeSi, as well as a blend of the two FeSi types. The blend was prepared by mixing 30% Cyclone 60 and 70% of Exxaro FeSi.

The operating conditions measured for all the tests are tabulated in Table 8. The D operating pressure values obtained during testwork ranged between 1 1 .7 and 1 3.8. Table 8: DMS operating conditions

Pilot DMS results

The results obtained during DMS pilot testwork for both the low grade and high grade samples are presented.

BIF (Low grade)

Table 9 below shows feed to the DMS reconstituted from the sinks and float products at different test conditions. The table shows that at different conditions, the content of hematite A in the feed was fairly consistent, ranging from 57.2% to 60.7%, with the bulk gangue contribution being quartz.

Table 9: Reconstituted feed to DMS using sinks and float PTA results for low grade sample

Table 10 summarises the results obtained, for testwork on the low grade sample. The D operating pressure values ranged between 1 2.8 and 1 3.8. The table attempts to give an indication of the reliability and reproducibility of the test results. Values of Fe grade that are coloured compare chemically assayed and recalculated head or sinks grades using PTA. The values in brackets are of directly measured feed samples and are not recalculated. It can be seen that the PTA agreed reasonably well with the measured data. Note all S1O2 and AI2O3 values reported are calculated using PTA.

For both low and high grade samples, tests done with no slimes present are denoted by the postscript N. No clearly outstanding trends with either FeSi type or with feed relative density can be easily seen from this view of the results other than:

1 . The test with no slimes present (Exxaro 3.8N) performed very poorly, showing that the presence of the slime fraction has a positive (stabilising) effect on cyclone performance. 2. The test at highest feed density (Blend 4.0) performed very poorly and did not produce a high grade sinks.

Table 1 0: DMS performance using low grade sample

Jig Fines Feed (High Grade) sample

Table 1 1 below shows feed to the DMS, reconstituted from the sinks and float product at different conditions for the high grade sample. The results indicate the significant presence of Flematite C, which will have an effect on the quality of product expected. Compared to the low grade material quartz is present as the predominant gangue mineral while the clay mineral content has increased to around 5%.

Table 1 1 : Reconstituted feed to DMS using sinks and float PTA results for high grade sample

Table 12 summarises the results obtained for testwork on the high grade sample. The D value calculated was around 1 1 .8 for all tests. The table indicates that the head grade for the sample was around 48%Fe, measured by chemical assay or via PTA.

The results show that four of the tests calculated higher head grades > 50% Fe (both PTA and assay), compared to the measured head feed of 48.2% Fe. All quality checks, resulted in similar results, with repeat testing also reporting elevated calculated headgrades. The discrepancy could thus not be explained.

Similar to the low grade results, no obviously discernible trends are noticeable from this view of the results, other than the clear poor performance of the run without slimes again.

Table 1 2: DMS performance using high grade sample

Modelling of DMS results

A least squares method was used to solve for the model parameters (Yp, Rhop, p q) for each individual run. Taken into account for the sum of squares minimisation were:

• DMS sinks and floats mass split

• DMS sinks and floats particle size distribution

• Fe grade distribution per size class in DMS sinks and floats

The sections below outline how well the model fit the DMS testwork results for both the low grade and high grade sample.

BIF (Low grade)

Table 13 shows the model parameters obtained to describe the DMS performance for the low grade sample. The results show that the actual cut-points for 1 mm particles obtained during the pilot DMS runs on low grade material for all conditions were much higher than expected for coarse particle DMS operation, in all cases even exceeding the DMS U/F density.

Table 13: Weibull DMS parameters and Ep (1 mm particles) for low grade sample

The modelled DMS partitions by size and density are plotted in Figure 1 2 to 1 6 for selected tests on the low grade BIF sample. All the modelling was done using the PTA data obtained on the product of each DMS test. Solid lines in both size and density partition graphs represent the PTA data as measured, while the dotted lines represent the best fit results of the Weibull model. Note that no stereological corrections have been made (size or liberation) during this Weibull model parameter optimisation phase.

Results show reasonably good fits between the measured and modelled cyclone performance for each of the different operating conditions. In all cases the DMS sinks showed particle size distributions showing preferential coarse particle recovery. Much less upgrading of finer size classes shown in the grade by size graphs where fine sinks and floats grades approach the fine feed head grade.

Jig Feed Fines (high grade) sample

Table 14 shows the model parameters to describe the DMS performance for the high grade sample tests. The same behaviour is observed as in the low grade sample, with the actual cut- points obtained during the pilot DMS runs for all conditions being much higher than the feed density.

Table 14: Weibull DMS parameters and Ep (1 mm particles) for high grade sample

The modelled DMS partitions by size and density are plotted in Figure 1 7 to 23 for selected tests on the high grade sample. Again reasonably good fits were obtained by the Weibull model, showing preferentially coarse sinks particle size distributions and preferential coarse hematite upgrading.

The partition by density graph in Figure 23 for the high grade sample at 3.8SG Exxaro with no slimes shows very poor performance with highly flattened partition curves for all size classes.

The results obtained showed that each individual DMS test can be quite well described using Weibull model with parameters (Y-p, Rho-p, p and q) that apply for the chosen operating conditions. These models can then be used to predictively simulate the response of different samples (mineralogical composition or particle size distribution).

Table 15 and Table 1 6 show the comparative efficiencies of each of the two FeSi types as well as the Blend case. Table 15 uses a 1 mm particle separating at a D50 of 4.50 as the base case and shows how both Ep and D50 for finer sized particles would separate. In all cases D50 for the finer size classes increases and Ep decrease.

Table 15: Comparative cyclone performance - with 1 mm @D50 of 4.50

Table 1 6 uses a 1 mm particle separating at a D50 of 3.20 as the base case and shows how both Ep and D50 for finer sized particles would separate. In comparison to Table 14, note that much more efficient separation across all size classes is evident, with quite similar Ep values for both Cyclone 60 and Exxaro fine FeSi.

Table 16: Comparative cyclone performance - with 1 mm @D50 of 3.20

Simulations

Three sets of DMS simulations were done by keeping the high and low grade feed compositions and size distributions constant while stepping through a range of DMS feed densities for each FeSi type. In this way, trends in performance can easily be compared:

1 . Using an average PTA size and compositions for the low and high grade samples respectively. This represents a finer particle size distribution and the grade measured by sieve analysis and XRF fusion.

2. Using screened particle size distributions for the low and high grade samples. This represents slightly coarser DMS feeds.

3. Higher Flematite A content with lower Flematite C content resulting in the same Fe content in the DMS feed (the high grade sample only). This represents the effect of improved feed density distribution to the DMS cyclone.

Low grade sample

Figure 24 and 25 shows the results comparing the grade / recovery profiles for the upgrading of the low grade sample using the different FeSi types. Both figures are very similar with the trends produced, giving the following findings: 1 . At high Fe recoveries to the sinks product, all the different FeSi types produced very similar results. Although the Exxaro FeSi is expected to be the most unstable at low feed FeSi densities, it appears as though the addition of the slimes fraction has assisted in stabilising both atomised FeSi types.

2. The run done without slimes using Exxaro FeSi has resulted in a very low grade product and is not recommended.

3. In contrast to traditional grade / recovery profiles where low recovery is associated with high product grade, all the fines DMS curves show a maximum grade. Increasing the DMS feed density beyond this point does not improve grade, rather both grade and recovery are compromised. This is probably due to the rapidly increasing viscosity of the FeSi/water mixture with increasing slurry density.

4. As expected, the Exxaro fine FeSi performs best at high feed densities, (low product yield), followed by the Blend and finally the Cyclone 60. This is probably due to the improved rheology that the gas atomising process imparts to the Exxaro FeSi.

5. The simulations using the coarser feed sample, (Figure 24), consistently show improved grades are obtained for all FeSi types and feed densities. This result is a consequence of improved cyclone performance via better Ep values for the coarser size classes.

High grade sample

Figures 26 to 28 show the results comparing the grade / recovery profiles for the upgrading of the high grade sample using the different FeSi types. All three figures are very similar with the trends produced, and are in agreement with those obtained for the low grade sample. Figure 28 shows the predicted DMS performance with increased Hematite A content and decreased Hematite C content. As expected, the results indicate further improvement in product grade at all operating conditions due to less near density material with the lower Hematite C content.

Conclusions

• Characterisation of feed ore is essential in understanding product quality and recoveries available, in this instance, specifically the effect of porosity and micro-inclusions.

• Results again indicated the sensitivity of DMS to feed size distribution, when operating in the <1 mm region. The optimum bottom size fraction for fines DMS should be carefully considered, especially in light of alternative technologies that may be able to complement the DMS in the finer size fractions.

• Testwork again showed an off-set between feed operating density and outpoint. Fines DMS obtains a much higher actual outpoint than what is expected from feed density (based on >1 mm performance). Much lower feed media densities are thus required to achieve high density outpoints.

• An increase in feed density does not result in a cleaner product, due to a decrease in separation efficiency at higher FeSi slurry viscosity.

• It is recommended that further lowering the feed density

(<3.4g/cm3) be considered. Both cylone 60 and milled 1 50D FeSi can be evaluated. This might result in further increasing product recovery, whilst maintaining Fe grade.

• At lower feed densities (up to about 3.6g/cm3), Cyclone 60 and Exxaro FeSi obtain similar results for both ores tested.

• The presence of slimes in the FeSi media was demonstrated to be essential in obtaining efficient separation. • DMS Fines model parameters were successfully established and predictive modelling can be considered for simulating:

• The effect of a variety of feed size distributions on performance,

• The effect of a variety of mineralogical composition on performance,

• Two stages of DMS can be simulated such as rougher and cleaner,

• Performance of any new ore, characterised by PTA, and

• Conditions required for any product requirement.

Table 17: Low grade reconstituted feed size distribution using

PTA

Table 1 8: Low grade feed size distribution using screening

Table 19: High grade reconstituted feed distribution using PTA

Table 20: High grade feed size distribution using screening

Table 21 : Chemical analysis for DMS products (low grade and high grade samples)

Additional tests

Table 22: Summary of DMS testwork for low grade and high grade samples

The method according to the present invention may be described further with reference to Figures 31 and 32 as follows.

PROCESS DESCRIPTION

Feed Preparation

Feed comprises of -1 mm crushed iron ore in one tonne bulk bags fed into a hopper with the use of a jib crane. Material is withdrawn from the hopper onto a conveyor belt which discharges material into a feed preparation screen mixing box where material is re-pulped prior to screening at 400pm.

Dense Medium Separation

Feed preparation screen oversize reports into a mixing box where it is mixed with the correct medium. The mixture of ore and medium is pumped to a 1 50mm dense medium cyclone where feed pressure is controlled using a variable speed drive fitted on the cyclone feed pump.

The dense medium cyclone underflow (sinks) (product) gravitates to the product magnetic separator via a mixing box where it is diluted with effluent water. The product magnetic separator effluent and non

magnetics report to the product dewatering/rinse screen fitted with 21 2 pm polyurethane panels. The product dewatering/rinse screen oversize is collected into one ton bulk bags for removal while the screen undersize reports to the effluent tank.

The dense medium cyclone overflow (floats) (discard) gravitates to a waste medium drainage screen fitted with 212 pm polyurethane panels. The waste medium drainage screen undersize reports to the pipe densifier feed tank. The waste medium drainage screen (2.1 m by 1 .08 m high frequency screen) oversize is re-pulped with effluent water prior to feeding the waste magnetic separator for FeSi recovery. Magnetic separator effluent and non-magnetics (discard) report to the waste dewatering/rinse screen to recover excess water from discard prior to disposal. The waste dewatering screen undersize reports to the effluent tank.

FeSi Recovery Circuit

Dense medium cyclone underflow is suitably diluted with effluent water ahead of the product magnetic separator (2000 Gauss CRM 60:60) for FeSi recovery. Flowever, the dense medium cyclone overflow reports to a waste medium drainage screen prior to feeding the waste magnetic separator (2000 Gauss CRM 60:1 00) for FeSi recovery. The overdense (magnetics) from the both the product and waste stream magnetic separators report to the circulating medium tank via a demagnetizing coil. The product magnetic separator underflow containing effluent and product report to a 2.9 m by 1 .08 m high frequency dewatering screen.

Waste magnetic separator recovers FeSi from the waste medium drainage screen oversize. Effluent and waste material from the waste magnetic separator report to the waste dewatering screen for excess water recovery prior to waste disposal. Effluent water reports to the effluent tank and used as dilution water to both product and discard magnetic separators. A portion of the effluent water is pumped to the effluent magnetic separator to scavenge any FeSi that might be in the effluent. The effluent magnetic separator underflow is used for fresh feed re-pulping and dilution ahead of the feed preparation screen.

Circulating Medium and Densification Circuit

A portion of the correct medium from the header box feeds to the circulating medium tank (3 m3) together with the magnetic separators overdense, densifying cyclone underflow, pipe densify underflow and product medium drainage screen underflow. The tank is fitted with an agitator to keep FeSi in suspension. The circulating medium tank is fitted with 2 pumps, namely, the circulating medium pump and primary densifier pump. The circulating medium pump discharges the correct medium at the medium headbox. A portion of the medium is distributed to the primary feed mixing box while a portion of the medium is recycled back to the circulation medium tank. The primary densifier cyclone pump feeds the densifying cyclone (1 50mm diameter). The primary densifier underflow gravitates back to the circulating medium while the overflow gravitates to the secondary pipe densifier feed tank. Pipe densifier overflow reports to the waste magnetic separator for FeSi recovery.

Density Control System

Circulating medium density control is achieved with a use of a DebTech density controller fitted with a density gauge mounted on the correct medium delivery pipe. The density controller is linked to an automatic valve for dilution water to the circulating medium tank.

Products / Discard Handling

The dewatered iron ore product and discard from the dewatering screens will be discharged into one tonne bags and removed with a use of pallets jacks and occasionally with the use of a forklift. Filtered slimes will also be removed with a use of a fork lift.

Slimes Handling

The feed preparation screen undersize gravitates to a desliming tank from where it is pumped to a geotube for dewatering. Dewatered solids together with a geotube will be removed using a forklift for storage and further processing if required. The geotube filtrate will gravitate to the geotube area sump from where it is pumped to the clarified water tank.

Dense Media Separation (DMS) Pilot Plant Work

In this work product, the following contractions have the following meanings:

CM - Correct Medium (Ferro Silicon mixed with water at the required ratios for Dense Medium Separation) KGT - Conglomeratic Ore Type

PE - Flagstone Ore Type

BIF - Banded Iron Formation Ore Type

ATS - Anglo Technical Solutions, Crown Mines

-1 mm DMS Circuit at 63.5% and 55%Fe

The DMS piloting was conducted using Exxaro Fine NGA FeSi at operating densities ranging from 3.2 up to 4.0 t/m ³ . Optimum operating DMS cyclone pressure (Cyc. Operating Pressure) was found to be at 1 5D. During commissioning, different operating pressures and FeSi types were tested until the plant was operating stably with optimal differentials for the fines between 25% and 6%.

Shale Mass Balance Results

One of the objectives for the DMS plant was to produce a fine product at 63.5% Fe in line with the current Sishen fine product specification. For the -1 +400pm, the DMS plant achieved the fine product specification at 3.4 operating density with 54.5% yield as per Table 223.

Table 23: DMS performance on Shale (-1+400pm and -1+200pm)

The DMS cyclone was also fed with Shale samples a bottom size of 200pm to determine the impact of size range as the breakaway size for the 1 50mm Diameter cyclone is around 200pm. Moving from a 400pm bottom size, to 200pm bottom size thus feeding the plant with -1 +400pm results in 29% more material in the feed for shale. As expected, the yield achieved from the -1 +200pm drops for the same operating conditions as compared to the feed at -1 +400pm due to poor separation of the fines. The impact gets worse as the operating density increases which is effectively the effect of viscosity on the fines.

The piloting work on Shale show that the -1 +400pm fraction can produce on specification fine product qualities at relatively low operating densities of 3.4 t/m ³ . The -1 mm DMS separation efficiencies are negatively affected by lowering the feed bottom size from 400pm to 200pm and makes it difficult for to achieve on specification fine product from the low grade Shale.

BIF Mass Balance Results

Similar to Shale, the DMS plant was to produce from BIF a fine product at 63.5% Fe in line with the current Sishen fine product specification as well as a 55% Fe for flotation. For the -1 +400pm, the DMS plant achieved the fine product at 63% Fe while operating at 3.6 t/m ³ density with 44% yield as per Table 24.

Table 24: DMS performance on BIF (-1+400pm and -1pm+200pm)

Feeding the DMS plant with BIF at a bottom size of 200pm effectively brings 27% more material in the feed. The yield impact as a result of dropping the bottom size for BIF was not obvious.

The piloting work on BIF show that the -1 +400pm fraction can produce on specification fine product qualities at relatively low operating densities of 3.2 t/m ³ for blending into the rest of the fine product.

KGT Mass Balance Results

For KGT, the DMS plant was operated to produce a fine product at 63.5% Fe in line with the current Sishen fine product specification as well as a 55% Fe for flotation. From the -1 +400pm, the DMS plant achieved the fine product at 63.8% Fe while operating at 3.6 t/m ³ density which resulted in 47% yield as per Table 25.

Table 25: DMS performance on KGT (-1 +400pm )

The effect of bottom size was not performed for KGT as the scope addition was included after KGT material was processed through the DMS plant.

PE Mass Balance Results

For PE, the DMS plant was operated to produce a fine product at 63.5%

Fe in line with the current Sishen fine product specification as well as a 55% Fe for flotation. From the -1 +400pm, the DMS plant could only make 60% Fe while operating at 3.6 t/m ³ density which resulted in 40% yield as per . Further increasing operating density to 3.8 t/m ³ did not help improve the product quality but significantly reduced the yield by almost 20% points. In this case the significant drop in yield is a result of PE’s poor liberation characteristics as well as increase viscosity at higher operating densities.

Table 26: DMS performance on PE (-1+400pm)

The impact of bottom size on PE was not studied further as the plant already struggled to make flotation feed with a narrower feed particle size distribution.

JIG Grit (-1+400mih) Mass Balance Results As part of the Low Grade -1 mm beneficiation technology development, the current Jig -1 mm was processed through the DMS pilot plant to evaluate the potential for producing on specification fines product which the current test module Reflux Classifier at Sishen battles to achieve for various reasons. Jig -1 mm received was deslimed at 200pm and processed as - 1 mm+200pm as well as -1 mm+400pm, results of which are summarised in Table 27.

Table 27: DMS performance on Jig Grit (-1 mm+400pm and -1mm +200pm )

The -1 mm DMS pilot plant was able to achieve 64.0% Fe and above from 3.4 t/m ³ operating medium density for the -1 +400pm grit as presented in Table 27. For the -1 +200pm grit, reducing the DMS cyclone feed bottom size from 400pm to 200pm has a negative impact on yield as would be expected due to poor recovery of material closer to the breakaway size. What is also noticeable is the significant drop in product yield from 3.6 t/m ³ operating densities and furthermore to 3.8 t/m ³ as illustrated in Figure 33.

The drop in product yield and grade when operating density is increased to 3.8 can be attributed to increased medium viscosity which for the fines is significant. Figure 34 shows the medium rheology for the Exxaro fine nitrogen gas atomised FeSi used at the DMS pilot plant.

From 3.8 t/m ³ operating density, the medium viscosity starts increasing exponentially which would lead to poor selectivity as the outward centrifugal forces on dense fines particles is no longer sufficient to overcome the hydrodynamic drag from the viscous medium.

DMS (-1 mm+400pm) Performance Relative to Ore Potential

Mineralogical analysis on the -1 mm+400pm feeding the DMS pilot plant was performed to determine the theoretical grade yield potential for each ore type as illustrated in Figure 35. The DMS pilot plant operated with reasonable separation efficiencies as measured by how close the piloting grade yield performance approached the theoretical grade yield curve. Scott Napier Munn model is further applied to assess the actual Ep (separation efficiency) and cut densities in the sections to follow.

The DMS pilot plant on -1 mm+400pm performance relative ore potential is similar to what was achieved on the Coarse Reflux classifier with the largest average difference between actual and theoretical yield grade curve at 1 5%. Metallurgical performance of the -1 mm DMS plant and - 1 mm Reflux classier were found to be relatively similar. The choice of which processing technology to apply on the -1 mm flowsheet will be dependent on the outcomes of a trade-off study considering Capex, Opex and operability amongst other elements.

DMS (-1 mm+400pm) Separation Efficiency Curves

The same approach was followed to model the DMS pilot plant

performance on -1 mm+400pm size fraction. The DMS cyclone feed, product and waste stream samples from selected runs were subjected to Mineralogical analysis using an MLA to determine the density distribution by size. The selected piloting run for modelling BIF -1 mm+400pm DMS cyclone separation efficiency as presented in Figure 36 came from the

performance where the pilot plant achieved 43% mass yield at 60% Fe Grade. From the partition curves, it could be determined that the effective cut density achieved was 4.30 t/m ³ at an overall Ep of 0.38.

The selected piloting run for modelling Shale -1 mm+400pm DMS cyclone separation efficiency as presented in Figure 36 came from the

performance where the pilot plant achieved 55% mass yield at 63% Fe Grade. From the partition curves, it could be determined that the effective cut density achieved was 4.52 t/m ³ at an overall Ep of 0.40.

The selected piloting run for modelling KGT -1 mm+400pm DMS cyclone separation efficiency as presented in Figure 36 came from the

performance where the pilot plant achieved 48% mass yield at 63% Fe Grade. From the partition curves, it could be determined that the effective cut density achieved was 4.40 t/m ³ at an overall Ep of 0.45.

For PE, the selected piloting run for modelling -1 mm+400pm DMS cyclone separation efficiency as presented in Figure 36 came from the

performance where the pilot plant achieved 41 % mass yield at 60.2% Fe Grade. From the partition curves, it could be determined that the effective cut density achieved was 4.65 t/m ³ at an overall Ep of 0.28.